LOCATIONS
24 → 42
▲ 75.0% · 18 opened inside the window
The story the book tells, backed by instruments a diligence team can re-run themselves. Every claim below carries the instrument that proves it, the procedure a third party uses to reproduce it, and — in the same size type, in the highlighted column — what it does not prove.
WHAT THIS DOCUMENT IS — AND WHAT IT DELIBERATELY IS NOT
This is not a narrative deck. It is a claim register. A growth story survives a data room only if every sentence in it can be taken apart by a stranger and put back together the same way. So each claim is written as four things: the claim, the instrument that produced it, the procedure a diligence analyst follows to reproduce it without us, and the boundary of what it proves. The fourth column is the one that makes the other three believable. Nothing here is written to be persuasive. It is written to be checkable, which turns out to be the more persuasive of the two.
HEADLINE VITALS · INTAKE (DAY 0) → DISCHARGE (DAY 1030) · FULL TABLE IN THE SEALED RECORD
LOCATIONS
24 → 42
▲ 75.0% · 18 opened inside the window
NEW PATIENTS / LOC / MO — PROGRAMME COHORT
34 → 47
▲ 38.2% · 16 of 42 locations
NEW PATIENTS / LOC / MO — ALL LOCATIONS
34 → 40
▲ 17.6% · the ramp is in this number
COST PER NEW PATIENT — ALL-IN
$268 → $292
▲ 9.0% — WORSE. We lead with this one.
THE NARRATIVE — SIX PARAGRAPHS, NO ADJECTIVES THAT AREN'T LOAD-BEARING
The group was admitted at 24 locations in January 2023, running 34 new patients per location per month. Henry Schein One's 2026 Catalyst Index puts the industry average at 39 and the top decile at 82, so the group entered below average and never approached the top decile. It is not a corporate chain. Nine partners hold equity across six regions and each region carries its own P&L. There is no head office that can order a media budget into a region — there is only a case you make to nine people who each own their own outcome and each get to say no.
That ownership structure decided more about this engagement than any campaign did. Three regions bought into the acquisition programme at the start, covering sixteen of the eventual forty-two locations. Two of the three regions that declined were right to decline: several of their sites were already at or near chair capacity, where added spend buys demand the schedule cannot absorb and the only result is a worse cost per patient. Sixteen of forty-two is an adoption number, not a scope we chose, and reporting it as scope would flatter us. What the holding company could mandate, it did — intake instrumentation and reporting went company-wide, because measurement is a shared service on a shared cost line while media is regional money.
The examination (Form 247-B) found two things worth the engagement. First, dispersion: a 7.1× production spread between pods running identical staffing. Second, a measured leak at the front desk — roughly three qualified new-patient calls per location per month died unbooked. Three calls × 24 locations × 12 months = 864 patients a year. Valued at the federal mean annual dental expenditure per person with a dental expense ($887, AHRQ MEPS Statistical Brief #555), that is about $766,000 a year, and the arithmetic is the whole claim.
Over thirty-four months, new patients per location per month rose 34 → 47 inside the programme cohort and 34 → 40 across all forty-two locations. Case acceptance moved 44% → 50% against Henry Schein One's 45% average and 75% top decile. New-patient wait fell 21 → 16 days. Retention at locations open eight years or more moved 57% → 64%, against a published 58% average. Production dispersion closed to 2.9× inside the cohort and 4.1× across the whole group.
What it cost is on the same page as what it earned. Cost per new patient fell to $214 inside the programme cohort and rose to $292 all-in. Collections per new patient over the first twelve months fell 6.6%, $1,040 → $971: the volume that arrived was less established and carried more payer discount than the base it joined. That is the standing cost of buying growth, not the exception to it. Eighteen of the forty-two locations were opened during the engagement, twelve of them de novo, six in towns where the brand had never traded.
A group that adds eighteen locations in thirty-four months and shows no decline in its blended per-location figures is describing something that does not happen. The blended decline in this file is a ramp. The cohort files in tab 06 of the diligence index (Form 884-D) let a buyer read the ramp separately from the mature base and decide for themselves. Whether the buyer accepts that reading is the buyer's decision, not ours — and a narrative that assumes the favourable reading is the one that gets taken apart in the first management meeting.
THE FOUR-COLUMN RULE — HOW EVERY CLAIM IN THIS DOCUMENT IS BUILT
A claim that cannot survive all four columns is not in this document. Four such claims exist. They are printed on sheet 4 under CLAIMS WE DECLINE TO MAKE rather than deleted, because the ones you refuse to make are how a buyer calibrates the ones you do. The third column is what we're actually selling.
Column three is the credibility engine. It is not a disclaimer and it is not legal cover — it is the part a diligence analyst reads first to decide whether to trust columns one and two.
INSTRUMENT
Form 247-B (Day-0 baseline, definitions frozen February 2023) plus thirty-two consecutive monthly board packs, Form 610-M. System of record: the group's practice-management system.
Frozen definition: a new patient is a first completed visit ever with the group. Not a lead, not a booking, not a call. The definition has not changed since it was frozen, and the freeze is time-stamped in tab 01.
HOW A DILIGENCE TEAM RE-RUNS IT
Take the PMS new-patient export by location by month, months 1–34 (data-room tab 03). Filter to the sixteen cohort location IDs listed in tab 01. Divide by sixteen and by month.
Roughly two hours in a spreadsheet. No ADMEN system is touched and no ADMEN file is required. The export is generated by a client employee with client credentials.
WHAT IT DOES NOT PROVE
Causation. There is no holdout. The programme was adopted by the regions that chose to adopt it, which is the definition of a self-selected sample.
It does not prove the other twenty-six locations would respond the same way — two of them are chair-capacity constrained and would not. And it does not prove the rate holds if the intake instrumentation stops. The rate and the instrument were built together.
INSTRUMENT
The same PMS export, unfiltered. Same frozen definition, same thirty-four months, same monthly packs.
Eighteen of the forty-two locations opened inside the window; twelve of those were de novo, six in towns carrying no prior brand recognition.
HOW A DILIGENCE TEAM RE-RUNS IT
Identical procedure to C-01 with all forty-two location IDs. The open-date register for every site is in tab 01, so the cohort can be re-cut any way the analyst prefers — by tenure, by region, by de novo versus acquired.
We publish our cut. We do not publish only our cut.
WHAT IT DOES NOT PROVE
That the blended figure measures the programme. It is a weighted average of a mature base and a ramp, and it is here so a buyer can see both — not so we can quote the higher of the two.
It also does not settle which number a buyer should underwrite. A buyer who marks the blended figure is entitled to; we would rather argue that in a management meeting with the cohort files open than discover it in a re-trade.
INSTRUMENT
The week-long call-centre sit in Rx 01, then continuous call recording and scoring under Form 247-K. Three qualified calls per location per month died unbooked × 24 locations × 12 months = 864.
Valued at $887 — AHRQ MEPS Statistical Brief #555, mean annual dental expenditure per person with a dental expense. The only federal per-patient dental spend series.
HOW A DILIGENCE TEAM RE-RUNS IT
Pull the retained recordings (tab 04, months 5–34) and re-score any two hundred against the published rubric. The rubric is one page — answered, offered, booked — and its definitions have not moved since month 6.
Scoring two hundred calls takes an analyst about a day. We have never asked anyone to take the leak figure on trust.
WHAT IT DOES NOT PROVE
That all 864 were recoverable, or that a recovered patient is worth $887 to this group. MEPS is a national per-person expenditure figure, not this group's collected value per new patient. It is a valuation convention chosen because it is public and checkable, not because it is precise. The group's own collected value per new patient over the first twelve months is $971, and that is the figure used wherever a decision was actually made.
Months 1–4 of recordings were destroyed by the vendor's 90-day default retention before we changed the setting. That quarter cannot be re-scored by anyone, including us.
INSTRUMENT
Monthly production per pod from the PMS, at comparable staffing, trailing three months. Pod roster and the staffing normalisation rule are in tab 01.
Published analogue for the pathology: MGMA DataDive across 141 multispecialty groups reports total medical revenue per FTE physician from $250K to $1.8M at comparable staffing — roughly a 7× spread.
HOW A DILIGENCE TEAM RE-RUNS IT
Production by pod by month (tab 03); compute max ÷ min on a trailing-three-month basis for the cohort and again for all forty-two. Both series are in the export; neither is computed by us.
WHAT IT DOES NOT PROVE
Attribution to the programme. Dispersion narrows under mean reversion alone, and thirty-four months is long enough for that to matter.
What the data supports is that it narrowed further and faster inside the cohort than outside it. That is a difference in differences on a self-selected sample. It is evidence, not proof, and we call it evidence. The MGMA figure is a benchmark for the disease, not a target for the cure — MGMA publishes no target dispersion and measures physician groups, not dental pods.
Two of the four claims on this sheet are weaker than they look, and both say so in their own third column. That is the design.
INSTRUMENT
The reconciliation strip on every Form 610-M board pack, months 3–34. Attributed gross production, less contractual adjustments (GL 4100), equals attributed collections, tied to GL 4000 with variance under 1%.
Standing rule, written into the pack: where attribution and the practice-management system disagree, the PMS wins. Months 1 and 2 predate the definitions freeze and are excluded rather than restated.
HOW A DILIGENCE TEAM RE-RUNS IT
Pick any month. Take the pack's attributed-collections figure, open the trial balance for the same month, tie it out. Thirty-two chances to catch us.
The signature on each pack belongs to a client employee, not to ADMEN. The tie-out was performed by the controller's team before it was published, not after it was questioned.
WHAT IT DOES NOT PROVE
Incrementality. Attribution here is last non-direct touch — a rule for assigning credit, not a method for establishing what would have happened without the spend.
We never ran a post-rollout holdout, and in a partner-equity structure we would not have been given one: no region will withhold spend from its own P&L to improve someone else's measurement. A buyer modelling incremental EBITDA from these numbers is modelling something we did not measure. That is the most important sentence in this document and it is deliberately printed on the claim it undercuts.
INSTRUMENT
Form 247-L loop minutes, months 7–34 — twenty-eight months, each with a named client owner and an attendance log. Eleven quarterly recommitment reviews (Form 610-Q), each carrying a pre-filled case for concluding the engagement.
Form 610-R monthly readout is opened and read-receipted; the open rate is reported in the pack, including the months it fell.
HOW A DILIGENCE TEAM RE-RUNS IT
Interview any desk manager with us out of the room. Ask who runs Thursday's huddle and what the last three flagged rows were. The correct answer is theirs, not ours.
Then read the eleven 610-Q memos. If an agency writes the case for firing itself four times a year and the client keeps renewing, that is a durable relationship rather than a captured one.
WHAT IT DOES NOT PROVE
Retention of those individuals through a transaction. Key-person risk sits with the operators who run the cadence, and a data room cannot fix that; only retention packages can.
It also does not prove the cadence survives a change of ownership. That has not happened yet. Post-close KPI retention through the first two quarters is the signal that would prove it, and it does not exist on the date of this document.
INSTRUMENT
PMS appointment lead time — first available slot for a new patient — recorded monthly across all locations.
Two published figures exist and they disagree: ADA Health Policy Institute's 2026 Survey of Dental Practice (Table 22) puts the average at 16.0 days for general practitioners; Henry Schein One puts it at 23 days with top performers near 7. We cite both.
HOW A DILIGENCE TEAM RE-RUNS IT
PMS lead-time export, tab 03, months 1–34, all locations. One query, one chart.
WHAT IT DOES NOT PROVE
That demand routing did it. Eighteen locations opened inside the window and added chairs shorten waits on their own arithmetic, with or without a marketing programme.
The cohort split in tab 06 narrows the question. It does not close it. Landing on the published ADA average is a reasonable place to be, not a distinction — and calling it one would be the kind of sentence that costs a management meeting its credibility.
INSTRUMENT
PMS patient-visit history for the 8+ year location cohort. Cohort definition and the retention window are frozen in tab 01.
Henry Schein One's 2026 Catalyst Index reports 58% average and 90% top decile on the same cut — so 64% clears the average and sits a long way below the top decile.
HOW A DILIGENCE TEAM RE-RUNS IT
Patient-level visit history export, tab 06. Apply the frozen window. The cohort is eleven locations and the file is small enough to check by hand.
WHAT IT DOES NOT PROVE
As much as it appears to. The denominator changed. A larger share of the patient base is newly acquired, and newly acquired patients carry shorter observed histories.
Some of this movement is composition rather than loyalty. We have not decomposed it, because the patient-level data needed to do it cleanly does not exist for months 1–6. We would rather say that than publish a decomposition we cannot defend.
A banker will ask for all four of these. The answer is no, and the reason is written down in advance so that it is a policy rather than an improvisation in a room full of buyers.
CLAIMS WE DECLINE TO MAKE — AND EXACTLY WHY
Each of these four was requested at least once during the engagement, in one form or another. The answer did not change between the request and the data room, which is the only reason it is worth anything now.
RE-RUN PROTOCOL — WHAT A THIRD PARTY NEEDS, AND WHAT THEY DO NOT NEED FROM US
| WHAT THE ANALYST NEEDED | SOURCE | ADMEN LOGIN REQUIRED? | CLAIMS SERVED | TIME |
|---|---|---|---|---|
| PMS exports — new patients, production, lead time, visit history | client PMS | no | C-01 C-02 C-04 C-07 C-08 | 14h 00m |
| Call recordings + the one-page scoring rubric | client call-tracking vendor | no | C-03 | 6h 40m |
| Trial balance + the 610-M reconciliation strips | client GL / controller | no | C-05 | 1h 05m |
| 247-L minutes, 610-Q memos, desk-manager interviews | client shared drive / client staff | no | C-06 | 3h 00m |
| Definitions register, cohort rosters, open-date register, staffing normalisation rule | tab 01 — authored by ADMEN, handed over at month 6, client-owned since | no | all eight | included above |
TOTAL 24H 55M ACROSS EIGHT CLAIMS · MEDIAN 2H 35M PER CLAIM · SLOWEST C-03 (RE-SCORING 200 CALLS) AND C-08 (PATIENT-LEVEL HISTORY), BOTH FLAGGED IN TAB 01 SO NOBODY DISCOVERS THEM UNDER TIME PRESSURE. ADMEN HOLDS NO DATA THIS CLIENT CANNOT REGENERATE, AND NO INSTRUMENT IN THIS DOCUMENT REQUIRES AN ADMEN LOGIN. IF THE FIRM VANISHED TONIGHT, THE DATA ROOM STILL CLOSES.
THE ONE THING THIS DOCUMENT IS FOR
A buyer is not paying for growth. A buyer is paying for growth they believe will continue after the seller leaves the building. Every column three on sheets 2 and 3 exists to narrow the gap between what we can show and what they will believe — by conceding the gap first, in writing, before anyone has to find it. Nothing in a data room is more expensive than a claim that falls over on day forty.
— specimen —
PARTNER, ADMEN · AUTHOR OF RECORD
— specimen —
REVIEW GATE — SECOND PARTNER, PRE-DELIVERY
— specimen —
CLIENT CFO · ACCEPTS THE REGISTER AS ACCURATE · DATE